The Chinese company Deepseek updated its artificial intelligence model R1, and the new version is published without the announcement on the HUGGING FACE platform.
The tests of the new model R1-0528 showed noticed progress in the ability to generate code.
According to the test that developed researchers from elite U.S. Uvivers, MIT and Cornel, the new model surpassed competitors such as Grok 3 mini.
Deepseek has not yet published technical details on improvements in the R1-0528 version, it is informally about “minor test update”, “Reuters” reports. Deepseek was expected to publish a new model of artificial intelligence R2 during May, but the launch was postponed without official reasoning.
Deepseek presented the model R1 in January and had already caused an alarm in the industry due to exceptional efficiency and low costs and knocked down the value of US technological companies.
His success showed that American export controls did not significantly slow down the development of artificial intelligence in China, which has launched rapid reactions of competition – including new models and prices as Openai and Google reports “B92”.
Follow us on our Facebook and Instagram page as well as on the X account.
(Vijesti.ba)
